[xmonad] UTF8 output from xmonad mangled

[Norbert Zeh]

Alexey Khudyakov [2010.04.28 1518 +0400]:
> On Wed, Apr 28, 2010 at 1:11 PM, Norbert Zeh <nzeh at cs.dal.ca> wrote:
> > Hi folks,
> >
> > I'm using xmobar as my status bar and have problems to get UTF8
> > characters to display correctly.  I did the following tests:
> >
> > 1.  Start xmobar, reading input from stdin.  Type in some unicode
> > characters.  Xmobar displays them fine.
> >
> > 2.  Add some unicode characters to my loghook in xmonad.  Things come
> > out garbled.
> >
> > 3.  Instead of piping to xmobar, make xmonad write to a file.  The file
> > contents are also garbled.  To be more precise, they are exactly what
> > xmobar shows me in the second test.
> >
> > So this seems like the problem is on xmonad's end, not xmobar's.  I'm
> > running the latest darcs version of xmonad.  Any ideas what could be
> > causing this?
> >
> What version of GHC do you use? Starting from version 6.12 GHC properly
> encodes strings. Earlier versions just truncated characters to 0-255 range.
> 
> If you use GHC 6.10 or earlier you should use encodeString from utf8-string
> package to encode string in utf8 before writing to pipe.
> 

Thanks Alex and Ivan,

before I say anything more on the issue, I need to do a bit more
tinkering to find out what's really wrong.  I'm running darcs xmonad and
darcs xmobar, compiled with GHC 6.12.1.  So the utf8 flag for xmobar
should have no effect.  (I tried with and without, and the result is the
same.)  What I've found out so far is

(1) xmonad seems to double-encode the utf string.  Namely, if I change
    my ppOutput to hPutStrLn handle . UTF8.decodeString, things come out
    *kind of* fine.

(2) By *kind of* in the previous point, I mean that characters like
    accented letters, which were mangled before, now show up just fine.
    However, some other characters, like arrows, come out messed up, and
    for this the problem seems to be at xmobar's end because the same
    happens if I pipe these characters into xmobar from an xterm.  In
    case, you're wondering: yes, the characters I'm trying to display are in
    the font I'm using.

I want to play around a little more to dig into the second issue, but
the first one is certainly strange.

Cheers,
Norbert